I'm trying to use Global Land Cover Characterization data from the USGS:
It is in the Interrupted Goode Homolosine projection. As I am working on a continental scale (South America) I need to reproject to WGS84. I've tried using the Goode Homolosine projection (land) in ArcGIS but the resulting raster was offset to the East by around 11km.
I shifted this using the header file of the original .bil file. I need to clip it down to the area of my study site and so used a snap[ raster of existing data (worldclim data). The snap did not work and my GLCC raster is still warped and does not line up with underlying rasters. I have seen the related questions on GIS stack exchange but none of those appear to work either. I'm running ArcGIS10 sp1.
Okay, here's what I did to get the forest cover Global Land Cover Characterization Data to line up with Worldclim data (for Latin America in WGS84 (geographic) projection).
Took the global file from the USGS website, added the header to the same folder and then projected to Goode Homolosine (land) in Arc Catalog.
Modified the Goode Homolosine projection to use the Sphere (Arc Info) as it's spheroid.
Clipped the bit I needed and projected it to WGS84 (Geographic).
I then resampled the GLCC data as it was at 1000m resolution whereas the worldclim stuff is in 30 arc seconds. Before resampling I checked all the environments and added a snap raster (the worldclim ascii I wanted it to match), extents, output coords, cell size, etc).
They now line up. Hacky, as I said but it worked. And just for the record, I'm working with MaxEnt (hence the ASCII requirement).
There is word on the wire that you need to turn off background geoprocessing in order to work around a bug in 10 with the snap raster option. Mine was off anyway. In addition, you can resample first and then add the snap raster to a "copy raster" command. Might be safer that doing it all in one.